Zhang - How Small Slip Surfaces Evolve Into Large Submarine Landslides Insight From 3D Numerical

RESEARCH ARTICLE How Small Slip Surfaces Evolve Into Large Submarine
10.1029/2022JF006640
Landslides—Insight From 3D Numerical Modeling
Key Points:
W. Zhang1,2 and A. M. Puzrin2
• W e revealed the physical mechanism
of submarine landslides initiated from 1
Department of Engineering, Durham University, Durham, UK, 2Institute for Geotechnical Engineering, ETH Zurich, Zurich,
small slip surfaces with different
Switzerland
patterns
• We simulated the entire 3D submarine
landslide evolution integrating the
pre-failure initiation and post-failure Abstract Submarine landslides are a major marine geohazard affecting resilience of offshore infrastructure
mass movement and coastal urban centers. Attention has previously been paid to quantifying post-failure dynamics of
• We quantified the slip surface growth
within a weak layer and its transition
catastrophic submarine debris flows and their consequences. However, pre-failure initiation and growth of
into slab failure above the weak layer a small slip surface evolving into a large submarine landslide are still less understood. This study aims to
explore the physical failure mechanism of submarine landslides initiated from small slip surfaces and to
Correspondence to:
quantify key features of failure evolution. They are achieved by modeling the entire three-dimensional (3D)
W. Zhang, landslide evolution, integrating the initiation and growth of slip surface, failure of slab above the slip surface,
wangcheng.zhang@durham.ac.uk post-failure mass movement and re-deposition of transported sediments, using a novel numerical method.
The characteristics of the slip surface growth within a favored layer and the patterns of the slab failure in
Citation: the overlying layer have been thoroughly discussed. The transition from the pre-failure slip surface growth
Zhang, W., & Puzrin, A. M. (2022). to the diverse post-failure mass movement are first observed and discussed with the 3D geometry effects,
How small slip surfaces evolve into revealing the complex cascading mass movement mechanisms. The criterion for unstable growth of a planar
large submarine landslides—Insight
from 3D numerical modeling. Journal slip surface and critical condition for slab failure are proposed. The findings from the study facilitate scientific
of Geophysical Research: Earth Surface, understanding of the evolution of historic events and help safeguarding offshore developments against
127, e2022JF006640. https://doi. submarine landslide recurrence.
org/10.1029/2022JF006640
Received 9 FEB 2022

Plain Language Summary A submarine landslide can impact a seafloor area as large as over
Accepted 7 JUL 2022 1,000 km 2 and trigger catastrophic tsunami waves, but originate from a relatively small initiation zone,
spreading within a “sandwiched” weak layer. The physical mechanism behind the failure extension is
Author Contributions: difficult to assess even with the help of state-of-the-art geophysical and geological investigations. To close
Conceptualization: W. Zhang the knowledge gap, we investigate the entire submarine landslide evolution from the pre-failure initiation to
Data curation: W. Zhang post-failure mass movement using a novel numerical method. From our modeling, we find that the catastrophic
Formal analysis: W. Zhang
growth of a slip surface along a weak layer, together with slab failure extension, could lead to very large
Investigation: W. Zhang, A. M. Puzrin
Methodology: W. Zhang submarine landslides. For example, upon certain conditions, a slip surface with an initial size of around only
Project Administration: A. M. Puzrin 100 m 2 can grow at considerable speeds and becomes as large as >100 km 2 within minutes. We also reveal key
Resources: A. M. Puzrin
features controlling the development of the slip surface growth and its transition into the slab failure, and hence
Software: W. Zhang
Supervision: A. M. Puzrin quantify different phases of failure evolution. The numerical replication of the submarine landslide evolution
Validation: W. Zhang improves the understanding of the complex cascading mass movement of submarine landslides, and can
Visualization: W. Zhang
facilitate safe marine and coastal developments, for example, for offshore renewables.
Writing – original draft: W. Zhang
Writing – review & editing: A. M.
Puzrin
1. Introduction
Submarine landslides are a major marine geohazard posing significant threats to offshore infrastructure and
marine animal habitats, and very large landslides may induce tsunami waves affecting resilience of coastal urban
centers. The deadly Papua New Guinea tsunami in 1998 likely caused by a post-earthquake submarine landslide
(Tappin et al., 1999) and the Hengchun Slide occurred after earthquakes in 2006, which broke at least nine cables
(Lin et al., 2010), are two of the catastrophic events occurred in the past several decades.
© 2022 The Authors. Layered marine sediments usually exhibit reduction in strength during shearing, because of the collapse of the
This is an open access article under
the terms of the Creative Commons inter-particle bonding structures and the accumulation of pore water pressure. The ratio between the initial (peak)
Attribution-NonCommercial License, and softened (residual) strengths of marine soil sediments is usually between 3 and 7, and such soil with strength
which permits use, distribution and sensitive to shearing is referred to as “sensitive soil” (Issler et al., 2015; L’Heureux et al., 2012; Skempton, 1985).
reproduction in any medium, provided the
original work is properly cited and is not For slope failure in marine sensitive soils, shearing failure within a basal slip surface might lead to the growth of
used for commercial purposes. the slip surface, eventually evolving into a large translational submarine landslide.
ZHANG AND PUZRIN 1 of 24

21699011, 2022, 7, Downloaded from https://agupubs.onlinelibrary.wiley.com/doi/10.1029/2022JF006640, Wiley Online Library on [26/10/2023]. See the Terms and Conditions (https://onlinelibrary.wiley.com/terms-and-conditions) on Wiley Online Library for rules of use; OA articles are governed by the applicable Creative Commons License
Journal of Geophysical Research: Earth Surface 10.1029/2022JF006640
Figure 1. Conceptual evolution of translational submarine landslides: (a) local slip surface initiation in weak layer by triggers such as seismic loading; (b) slip surface
growth after triggers; (c) failure of slab above weak layer; (d) post-failure mass movement.
Figure 1 shows a conceptual evolution of typical translational submarine landslides. Although their scales are
usually very large, for example, the Storegga Slide involved a total volume of 3,500 km 3 of sediment debris
(Haflidason et al., 2004; Micallef et al., 2007), submarine slides might be initiated at a minute slip surface, as
shown in Figure 1a, triggered by external factors such as earthquakes. An initial slip surface is often concentrated
in a favored soil layer (the so-called ‘weak layer”) where shear strength relative to the overburden pressure is
lower than adjacent layers. The weak layer provides a locus for progressive growth of the slip surface with exter-
nal triggers (Locat et al., 2014), as shown in Figure 1b. Once the size of the slip surface reaches a threshold, the
slip surface growth becomes catastrophic and can only be limited by slope flattening or slab failure above the
slip surface (Puzrin et al., 2015, see Figure 1c). Diverse post-failure behavior, such as retrogressive failure with
the main rupture surface extending along the upslope direction (Kvalstad et al., 2005) and progressive plow-
ing with continuing erosion of sediments from the seabed by the frontally confined slide mass (Frey-Martínez
et al., 2006; Zhang et al., 2021) may evolve after the slab failure, as illustrated in Figure 1d. However, the mech-
anism of enormous translational landslides evolving from insignificant initial failure, particularly under complex
three-dimensional (3D) terrains, remains less understood, which impedes the scientific understanding of historic
events and quantification of submarine landslide risks in the future.
Submarine slope stability analysis and assessment of landslide dynamics and impact to offshore infrastructure are
important areas of study for engineering geology professionals and geotechnical engineers. Previous studies have
treated the pre-failure initiation and post-failure flow dynamics of submarine landslides separately, and no clear
vision has been given to the evolution from pre-failure to post-failure phases. Slope stability (failure initiation) is
usually simplified as a two-dimensional (2D) plane strain problem (Cornforth, 2005; Morgenstern & Price, 1965;
Spencer, 1967), which has been considered conservative compared to the 3D treatment in the Limit Equilibrium
Method (LEM), as the resistance in the out-of-plane direction is neglected. However, the slip surface of transla-
tional landslides can grow in any direction within a favored “weak” layer, which may or may not be parallel to the
main travel direction of the slide mass as depicted in Figure 1. Such a multi-directional propagation mechanism
can provide additional driving force because of the reduction in strength during slip surface growth, but this
physical failure mechanism cannot be considered in a 2D scenario. A few LEM studies have been able to consider
3D submarine slope stability analysis with rotational sliding mechanisms in the past two decades (Somphong

et al., 2022; Sultan et al., 2007). Zhang et al. (2020) and Klein and Puzrin (2021) studied the instability of 3D
submarine slopes with translational sliding mechanisms ignoring inertia effects and considering simplified planar
and conical slope geometries, respectively. The robustness of the proposed criteria and their applications in more
general conditions need to be validated.
Simulation of the slide mass movement after slab failure has been performed by using sophisticated large defor-
mation numerical methods such as computational fluid dynamics (Biscarini, 2010), smoothed particle hydrody-
namics (Zhang & Randolph, 2020) or the material point method (Dong et al., 2017), which are, however, limited
to 2D small scale modeling due to computational inefficiency. The depth-integrated method based on the shallow
water condition is the most efficient numerical method for the debris flow dynamics modeling pioneered by
Imran et al. (2001) and advanced by many following studies (De Blasio et al., 2004; Elverhøi et al., 2005). Some
3D numerical programmes such as Bingclaw (Kim et al., 2019), based on the depth-integrated method, have been
emerging recently for modeling dynamics of both slide mass and generated tsunami waves. However, they require
prior knowledge of details of the initial slide mass such as geometry, volume and velocity, which are rarely deter-
mined in practice with sufficient accuracy. A correlation between slope stability analysis (landslide initiation)
and evolved debris flow (landslide dynamics) has been given for a 2D cross-section of a submarine slope (Zhang
& Puzrin, 2021). This is yet to be extended to the 3D scenario considering the multidirectional expansion which
is essential for fully understanding the failure evolution and more importantly for quantifying impact area and
involved volume of a submarine landslide.
Therefore, the main aims and novelty of this study are to reveal how a small slip surface evolves in a 3D geometry
(with lateral expansion) into a large submarine landslide, and to quantify its different phases of failure evolu-
tion. They are achieved by integrated modeling of pre-failure initiation and post-failure behavior of submarine
landslides with 3D geometry effects using a novel numerical method. This numerical method is tested against
a historic event occurred on the eastern margin of New Zealand. Key features and controlling factors for the
transition from multidirectional slip surface growth within the weak layer to post-failure mass movement are
revealed. New criteria for slip surface growth are proposed and characteristics of slab failure are discussed. The
findings from the study facilitate scientific understanding of the evolution of historic events and help safeguard-
ing offshore developments against submarine landslide risks.
2. Methods
2.1. Slope Geometry
Four types of submarine slopes (planar, S-shape, convex, and concave, as shown in Figure 2) are used in the
study. Each of them consists of a submarine slope, an upper basin floor, and a lower basin floor from nearshore to
deep sea. The seabed profile is built of an overlying layer, a weak layer, and a base (undisturbed or less disturbed
layers below the weak layer). An initial slip surface was assumed to occur within the weak layer at the center
of the slope. The submarine slope is assumed to have an average inclination of 6° to horizontal. Glide planes
of submarine landslides are typically up to 10° on the upper slope, reducing to less than 5° on the lower slope
(Masson et al., 2006). The S-shape slope used in present study reflects this general trend reducing the slope angle
from 9° to nearly 0°. Investigations of slip surface growth and slab failure initiation focus on the planar slope,
which is assumed sufficiently long (8,000 m) and wide (6,000 m). The complete landslide evolution, including
the post-failure behavior and the arrest of mass transport deposit (MTD), is then simulated with considerations of
the full slope model and different 3D slope geometries.
A local Cartesian coordinate system x-y-z is used, with the origin set at the center of the initial slip surface. The
x–y plane (z = 0) was set as the horizontal plane and the z-axis points away from the seabed as shown in Figure 2.
The expression of the planar slope geometry is straightforward with the coordinate z linearly varying from the
slope crest to the toe (see Figure 2a). To describe the S-shape slope geometry, a curvilinear slope model was used,
where the weak layer is parallel to the slope surface and antisymmetric about the slope center. The slope center is
set as the origin of the coordinate system, and the weak layer geometry is described by (see Figure 2b)

Figure 2. 3D submarine slope models used in the study: (a) planar; (b) S-shape; (c) convex; and (d) concave.
⎧ [ ( 𝑦𝑦 )]
⎪−𝐻𝐻 1 − 𝑒𝑒𝑒𝑒𝑒𝑒 𝐻𝐻 𝑡𝑡𝑡𝑡𝑡𝑡 𝑡𝑡𝑐𝑐 , 𝑦𝑦 𝑦 0
(1)
𝑧𝑧 = ⎨ [ ( 𝑦𝑦 )]
⎪𝐻𝐻 1 − 𝑒𝑒𝑒𝑒𝑒𝑒 − 𝑡𝑡𝑡𝑡𝑡𝑡 𝑡𝑡𝑐𝑐 , 𝑦𝑦 ≥ 0
⎩ 𝐻𝐻
where
𝐴𝐴 𝐴𝐴c is the maximum slope angle at the center,
𝐴𝐴 and 𝐴𝐴 is the half-height of the slope. The 3D model is
constructed by extending the above function through the x axis.
The convex and concave slope geometries are constructed from a truncated cone and expressed, in terms of a
global Cartesian coordinate system X-Y-Z as shown in Figures 2c and 2d, by
( √ )
𝑋𝑋 2 + 𝑌𝑌 2 − 𝑅𝑅𝑡𝑡
(2) 𝑍𝑍 = 2𝐻𝐻 1 −
𝑅𝑅𝑏𝑏 − 𝑅𝑅𝑡𝑡
and
√
𝑋𝑋 2 + 𝑌𝑌 2 − 𝑅𝑅𝑏𝑏
(3)
𝑍𝑍 = 2𝐻𝐻
𝑅𝑅𝑡𝑡 − 𝑅𝑅𝑏𝑏
where 𝐴𝐴t and 𝐴𝐴b are radii of circular cross sections of the truncated cone through the slope crest and the slope toe,
𝐴𝐴
respectively. The local coordinates can then be easily related to the global coordinates based on the origin shifting
as shown in Figures 2c and 2d.
2.2. Governing Equations
Key assumptions for establishing governing equations are as follows.

• he thickness of the landslide is small (<1:10) compared to its dimensions, so that the velocity can be aver-
T
aged along the depth of each cell.
• Momentum of the slide mass along the z-direction is negligible.

• rapped water moves together with soils in each cell, and any generated pore pressures have no time to dissi-
T
pate, ensuring an undrained (and incompressible) condition.
Based on these assumptions, conservation of mass in each cell can be expressed by
𝜕𝜕𝜕 𝜕𝜕𝜕𝜕𝜕 𝜕𝜕𝜕𝜕𝜕

(4) + + =0
𝜕𝜕𝜕𝜕 𝜕𝜕𝜕𝜕 𝜕𝜕𝜕𝜕
where 𝐴 is the thickness of the slab above the weak𝐴𝐴layer,𝐴𝐴𝐴𝐴 and 𝐴𝐴 are the velocity𝐴𝐴in the 𝐴𝐴
𝐴𝐴 𝐴𝐴- and 𝐴𝐴-directions, respec-
tively,
𝐴𝐴 and 𝐴𝐴 is the elapsed time. Conservation of momentum in each cell is given by
𝜕𝜕𝜕𝜕𝜕 𝜕𝜕𝜕𝜕𝜕2 𝜕𝜕𝜕𝜕𝜕𝑥𝑥 𝜕𝜕𝜕𝜕𝜕𝜕𝜕 𝜕𝜕𝜕𝜕𝜕𝑥𝑥𝑥𝑥 𝜏𝜏𝑤𝑤𝑤𝑤𝑤 + 𝜏𝜏𝑔𝑔𝑔𝑔𝑔 + 𝜏𝜏𝑑𝑑𝑑𝑑𝑑𝑑𝑑𝑑𝑑𝑑𝑑

(5) + + + − − =0
𝜕𝜕𝜕𝜕 𝜕𝜕𝜕𝜕 𝜌𝜌𝜌𝜌𝜌𝜌 𝜕𝜕𝜕𝜕 𝜌𝜌𝜌𝜌𝜌𝜌 𝜌𝜌
and
𝜕𝜕𝜕𝜕𝜕 𝜕𝜕𝜕𝜕𝜕2 𝜕𝜕𝜕𝜕𝜕𝑦𝑦 𝜕𝜕𝜕𝜕𝜕𝜕𝜕 𝜕𝜕𝜕𝜕𝜕𝑥𝑥𝑥𝑥 𝜏𝜏𝑤𝑤𝑤𝑤𝑤 + 𝜏𝜏𝑔𝑔𝑔𝑔𝑔 + 𝜏𝜏𝑑𝑑𝑑𝑑𝑑𝑑𝑑𝑑𝑑𝑑𝑑

(6) + + + − − =0
for
𝐴𝐴 the 𝐴𝐴𝐴𝐴- and 𝐴𝐴-directions, respectively. In the above equations,
𝐴𝐴 𝐴𝐴 is the bulk density of saturated 𝐴𝐴 soil,
𝐴𝐴 𝐴𝐴𝑥𝑥, 𝐴𝐴𝑦𝑦 and
𝐴𝐴 𝐴𝐴𝑥𝑥𝑥𝑥 are stress components applied at the center of the cell face, with the face normals parallel 𝐴𝐴 to the
𝐴𝐴 𝐴𝐴 or 𝐴𝐴 axis;
𝐴𝐴 𝐴𝐴 and 𝐴𝐴w,𝑦𝑦 are weak layer (or slip surface) shear stress components;
𝐴𝐴w,𝑥𝑥 𝐴𝐴 𝐴𝐴 and 𝐴𝐴g,𝑦𝑦 are shear stress components at
𝐴𝐴g,𝑥𝑥
the buried depth of the weak layer by gravity; 𝐴𝐴 and 𝐴𝐴drag,𝑥𝑥
𝐴𝐴 and 𝐴𝐴drag,𝑦𝑦 are shear stress components by drag of ambient
water.
The governing equations are solved through a finite volume scheme with a staggered mesh strategy, which is
detailed in Appendix A. Determination and numerical implementation of marine sediment properties and drag
forces from the ambient sea water are also fully addressed in Appendix A, with key instructions given as follows.
Water entrainment into the slide mass and runout hydroplaning have not been considered at this stage in the study.
2.2.1. Marine Sediment Properties and Drag Force
The soil properties, such as stress and strength, in the weak layer and sliding layer are updated based on the
current values
𝐴𝐴 𝐴𝐴of 𝐴,𝐴𝐴𝐴𝐴 and 𝐴𝐴. Within the slip surface, the shear stress
𝐴𝐴 (𝐴𝐴w) is limited to the current shear strength,
which is reduced during shearing, and given by
( )
𝛿𝑝 1
(7) 𝜏 𝑤 = 𝑚𝑎𝑥 1− 𝑝, ⋅ 𝑠𝑢𝑤,𝑝
𝛿𝑟 𝑆𝑡
where 𝐴𝐴 p = ∫0 ‖𝜹𝜹̇ ‖𝑑𝑑𝑑𝑑 is the accumulated plastic shear displacement across the weak 𝐴𝐴layer, 𝐴𝐴r the value
𝐴𝐴 of 𝐴𝐴 p at
𝑡𝑡 p p
𝐴𝐴
the residual shear stress,
𝐴𝐴 𝐴𝐴t the soil sensitivity defining the ratio of the peak and residual shear strengths, and
𝐴𝐴 𝐴𝐴uw,p the peak undrained shear strength in the weak layer. Hydrodynamic pressure drag for a streamlined body
like a submarine sliding mass is less significant than the skin friction drag, and the latter can be approximated by
(Elverhoi et al., 2005; Norem et al., 1990)
( )
1 𝐿𝐿 −2.5
(8)
𝜏𝜏𝑑𝑑𝑑𝑑𝑑𝑑𝑑𝑑 = 𝐶𝐶𝑓𝑓 𝜌𝜌𝑤𝑤 𝑣𝑣2 ; 𝐶𝐶𝑓𝑓 = 1.89 + 1.62 𝑙𝑙𝑙𝑙𝑙𝑙
2 𝑘𝑘
where 𝐴𝐴f is the frictional drag coefficient,

𝐴𝐴 𝐴𝐴 𝐴𝐴w is the seawater density,
𝐴𝐴 𝐴𝐴 is the maximum length of the sliding mass
along the travel direction (i.e., y-direction)
𝐴𝐴 and 𝐴𝐴 is the roughness length of the sliding mass surface in the range
of 0.01–0.1 m (Norem et al., 1990). For a length of the sliding mass varying between 10 and 1,000 m, the friction
drag coefficient falls in the range of 0.005–0.016 according to Equation 8. In the present study, the value 𝐴𝐴 of 𝐴𝐴
was set to the upper limit 0.1 m, and the value
𝐴𝐴 of 𝐴𝐴 is measured from the numerical modeling results. It increases
during the submarine landslide evolution.
2.3. Verification
In this section, the developed numerical scheme is tested by modeling a realistic submarine landslide occurred on
the eastern continental margin of New Zealand. The landslide was recently discovered and reported by Watson

Figure 3. Verification of computational method by modeling a submarine landslide on eastern margin of New Zealand (Watson et al., 2020): (a) top view morphology
of the landslide for headscarp, lateral ridge and s; and (b) post-failure slope surface along the cross-section A-A′.
et al. (2020) with clear morphologies of head scarp, lateral ridges and MTD as shown in Figure 3a. The main
headscarp is observed around 3,000 m in width and at a water depth of 1,500 m, and the total runout distance
from the headscarp to the MTD front is about 10,000 m with the MTD found at a water depth of 3,500 m. The
pre-failure slope was reconstructed using the S-shape function with the maximum slope𝐴𝐴angle 𝐴𝐴𝑐𝑐 = 25◦ and the
half-height of the 𝐴𝐴slope 𝐴𝐴 = 1, 250 m, as shown in Figure 3b. Read from the bathymetry map, the depth of
the slip surface,
𝐴𝐴 𝐴, is about 100–200 m and was chosen as 150 m in the study. The peak undrained shear strength
of the sediments was set to 500 kPa for both the weak 𝐴𝐴 layer (𝐴𝐴uw,p) and the sliding 𝐴𝐴
layer (𝐴𝐴us,p), and the bulk
density was𝐴𝐴set to 𝐴𝐴 = 1, 800 kg/m 3. The soil sensitivity
𝐴𝐴 is 𝐴𝐴𝑡𝑡 = 2, and the parameter
𝐴𝐴 𝐴𝐴𝑟𝑟𝑝𝑝 is 0.2 m. The setting of
the parameters is to ensure the pre-failure slope is in a marginal state with the undrained shear strength (500 kPa)
slightly larger than the shear stress at the steepest point of the slope (497 kPa). The triggering factor of the event is
unknown but is likely related to significant seismic activity in the region. Nevertheless, it is reasonable to assume
that the failure starts from a small region at the steepest segment of the slope. Here, we assumed a circular initi-
ation zone of 200 m in diameter at the slope center.
The numerical results in terms of the final configurations of the headscarp, lateral ridges and MTDs are visu-
alized in Figure 3a. Figure 3b compares the post-failure seafloor surfaces from the numerical modeling and the
bathymetry map adapted from Watson et al. (2020). Compared to the observations from the bathymetry map, the
total runout distance is about 10% larger and the headscarp is slightly more extended. The numerical morphology
of the landslide is symmetric while it is not the case observed from the bathymetry map, since numerical mode-
ling ignored the heterogeneity of the slope geometry and of the sediment properties along the x axis. Despite
that, the results from the numerical modeling compare well with the realistic observations, thus validating the
applicability of the proposed numerical scheme to modeling of this 3D large-scale submarine landslide. The
mechanisms controlling the propagation of the slip surface along the weak layer, as well as the slab failure and
post-failure evolution of submarine landslides similar to the one simulated above will be demonstrated in the
remaining of the study.
3. Results
In this section, the numerical results of the whole evolution of submarine landslides are presented covering the
initiation and growth of slip surface, slab failure, post-failure evolution and final redeposition of slide mass.

Figure 4. A typical case with stable growth of slip surface: (a) shear strength contour; (b) distributions of shear strength along cross sections parallel to x-axis; and
(c) distributions of shear strength along cross sections parallel to y-axis. Three zones are identified during slip surface growth: slip surface with residual strength (red);
intact zone with peak strength (blue); and process zone with reducing strength (band color).
3.1. Initiation of Slip Surface
The initial local slip surface occurs where the permanent or transient driving force exceeds the resistance. It
forms either because of an increase in the driving force, for example, by seismic events or increase in slope from
diapirism, or a decrease in shear strength, for example, by soil degradation or accumulation of pore pressures.
The size and shape of initial slip surface depend on the terrain and triggering factors. In this initial study, it is
sufficient to assume the initial slip surface being symmetric with respect to both x- and y-axis, and its boundary
can be described by a series of functions:
| 2𝑦 |
𝑛
| 2𝑥 |
𝑛
| | +| | =1
(9)
| 𝑙𝑥 | | 𝑙𝑦 |
where 𝐴𝐴 is a shape parameter,

𝐴𝐴 𝐴𝐴 and 𝐴𝐴
𝐴𝐴𝑥𝑥 and 𝐴𝐴𝑦𝑦 are dimensions of the slip surface in the x- and y- directions, respec-
tively. The value
𝐴𝐴 of 𝐴𝐴 is larger than unity for a convex slip surface, and for most cases studied here, it was taken
as 2 representing for an elliptical or circular slip surface. The sediments within the slip surface are degraded by
slip weakening, time weakening or both, although the process may take a long time. Therefore, it is sufficient and
conservative to assume that the shear strength in the slip surface has been reduced to the residual.
The force imbalance from the slip surface is transferred and sustained by surrounding soils, which may undergo
plastic failure and post-peak strain softening. Figure 4 shows the contour of shear strength after the formation of
a circular slip surface
𝐴𝐴 (𝐴𝐴 = 𝐴𝐴
2, and 𝐴𝐴𝑥𝑥 = 𝑙𝑙𝑦𝑦 = 40 m), as simulated by using the proposed numerical scheme. At the
initial state, the shear strength is maintained at the peak value outside the slip surface. The at-rest lateral earth
pressure coefficient, which is the ratio between the horizontal and vertical earth pressures, was𝐴𝐴set to 𝐴𝐴0 = 0.5.
Other properties of the numerical model and materials are listed in Table 1. Within the weak layer, three zones can
be identified as shown in Figure 4a: the intact zone (where soils remain intact), the “process zone” (where soils
undergo strain softening, and shear strength ranges between the peak and the residual) and the slip surface (where
soils reach the residual state). Six cross sections, three in each (x- or y-) direction, are chosen to further observe
the distribution of the shear strength within the three zones, as given in Figures 4b and 4c. For the x-I, x-II, y-I and

Table 1 y-II sections which cross all three zones, discontinuities in the distributions
Base Parameters for Numerical Cases of the shear strength exist, dropping from a post-peak value to the residual.
These distinguish the process zone from the initial slip surface. The x-III and
Parameter Value Unit
y-III profiles, however, cross the process zone and intact zone only, and the
Overall model length,
𝐴𝐴 𝐴𝐴 4,000 m discontinuity in shear strength distribution is absent. Note that, in this case,
Overall model 𝐴𝐴
width, 𝐴𝐴 150 m the process zone is developed to fully resist the unbalanced forces from the
Slope𝐴𝐴angle, 𝐴𝐴 6.0 degrees slip surface, and hence the slope remains stable. This is defined as stable slip
Sliding layer thickness,
𝐴𝐴 𝐴 8.0 m surface growth and will be detailed in the next sub-section.
Shear stiffness in weak𝐴𝐴layer, 𝐴𝐴 1,656 kPa/m
Shear modulus in sliding𝐴𝐴layer, 𝐴𝐴 500 kPa 3.2. Stable Growth of Slip Surface
Peak shear strength in weak𝐴𝐴layer, 𝐴𝐴𝑢𝑢𝑢𝑢𝑢𝑢𝑢 10 kPa Figure 5 shows the development of the process zone during stable growth of
Residual shear strength in weak𝐴𝐴layer, 𝐴𝐴𝑢𝑢𝑢𝑢𝑢𝑢𝑢 2 kPa slip surface. For a relatively large slip surface𝐴𝐴(i.e., 𝐴𝐴𝑥𝑥 = 40 m for a circular
Plastic shear displacement to the residual strength,
𝐴𝐴 𝑝𝑝
𝐴𝐴𝑟𝑟 0.2 m slip surface), it can disturb and weaken adjacent soils, forming a significant
Plastic shear strain to the residual strength,
𝐴𝐴 𝑝𝑝
𝐴𝐴𝑠𝑠𝑠𝑠𝑠 0.2 process zone surrounding the slip surface. However, if the slip surface is
sufficiently small𝐴𝐴(e.g., 𝐴𝐴𝑥𝑥 = 20 m), the driving force from the slip surface
At-rest lateral earth pressure coefficient,
𝐴𝐴 𝐴𝐴0 0.5
might be easily sustained by the surrounding soils without the formation of
Characteristic length𝐴𝐴 a, 𝐴𝐴𝑐𝑐 10 m
the process zone, that is, the soils remain intact as shown in Figure 5. When
Submerged soil density,𝐴𝐴 𝐴𝐴 740 kg/m 3 the diameter of the slip surface grows 𝐴𝐴 to 𝐴𝐴𝑥𝑥 = 30 m, the process zone appears
√
p only at the front and rear of the slip surface. The above observation reveals
a𝐴𝐴c = .
𝐺𝐺𝐺𝐺𝐺r
𝐴𝐴 𝜏𝜏p −𝜏𝜏r that: (a) for a circular slip surface, the process zone first emerges at the rear
and in front of the slip surface; and (b) the larger the slip surface, the more
significant the process zone. Figure 5 also compares the shear stress contours
resulting from different sizes of slip surface. The shear stress is maintained at the value under gravity far away
from the slip surface
𝐴𝐴 (≈6 kPa for parameters listed in Table 1) and increases to the peak at the interface between
the intact and process zones. It is limited to the shear strength within the process zone and the slip surface.
Stability can be eventually achieved for the cases

𝐴𝐴 of 𝐴𝐴𝑥𝑥 = 20, 30 and 40 m despite the development of the process
zone. As defined above, the process of extensive expansion of the slip surface𝐴𝐴up to 𝐴𝐴𝑥𝑥 = 40 m can be termed
stable slip surface growth. In contrast, for the case
𝐴𝐴 of 𝐴𝐴𝑥𝑥 = 50 m, the growth of the slip surface cannot be restricted
under existing forces and hence is termed unstable slip surface growth. Note that Figure 5 shows a transient
Figure 5. Investigation of stable slip surface growth in terms of the evolutions of shear strength (top row) and shear stress
(bottom row) contours. The submarine slope remains stable with the growth of the slip surface up to 40 m in diameter and
turns unstable when the diameter reaches 50 m (the last column subfigures show a transient moment).

Figure 6. Investigation of unstable slip surface growth from initial diameter of 50 m in terms of evolutions of: (a) shear strength contours, (b) lateral velocity contours
and (c) vertical velocity contours for the case of free movement in x direction (perpendicular to the travel direction); (d) shear strength contours for the case of restricted
movement in x direction; and (e) a comparison of “peanut” and “ellipse” slip surface growth patterns. (a–c) present information for the case of free movement in x
direction showing a “peanut” pattern with more significant lateral expansion at shoulders of the slip surface; and (d) with restricted movement in x direction presents a
“ellipse” mechanism.
moment of this case, and the dynamic unstable expansion of the slip surface during unstable slip surface growth
will be discussed in the next section. Therefore, in this regard, one may imagine that a circular slip surface
gradually expands due to different processes over geologic time until its radius reaches 50 m causing a sudden
catastrophic expansion and resulting in a submarine landslide.
3.3. Unstable Growth of Slip Surface

With an initial slip surface
𝐴𝐴 of 𝐴𝐴𝑥𝑥 = 𝑙𝑙𝑦𝑦 = 50 m and other properties listed in Table 1, the growth of the slip surface
is unstable and can only be limited by slope flattening or slab failure. Figures 6a–6c, respectively, shows the
evolution of the shear strength contour, horizontal velocity field and vertical velocity field during the unstable
growth of the slip surface. To investigate the slip surface growth, the shear strength in the overlying layer was
intentionally set to a high value (1,000 kPa) to avoid slab failure.
𝐴𝐴 At 𝐴𝐴 = 10s, the slip surface grows from a circle to
an ellipse with the major axis parallel to the potential travel direction. Thereafter, the slip surface grows dramat-
ically and propagates more outward at the four shoulders, forming a distinctive “peanut” shape 𝐴𝐴 at 𝐴𝐴 = 25s. The
wide shoulders are generated because of larger horizontal velocity in these areas, whereas along the major (y-)
and minor (x-) axes of the slip surface, the horizontal velocity is close to zero, as shown in Figure 6b. The slip
surface is symmetric in terms of both x- and y-axes 𝐴𝐴 before 𝐴𝐴 = 50
𝐴𝐴s. At 𝐴𝐴 = 50s, however, the downslope part of the
slip surface is slightly larger than the upslope, as soil begins to be accumulated more downslope.

Figure 7. Investigation of slab failure in terms of the evolutions of contours of shear strength in weak layer (top row) and
plastic shear strain in sliding layer (bottom row) for the case
𝐴𝐴 with 𝐴𝐴us,p ∕𝑠𝑠uw,p =𝐴𝐴1 and 𝐴𝐴0 = 0.5. Note: after slab failure, the slip
surface does not grow as a complete “peanut”’ with only the lower half developing; the upper half is limited by the headscarp
and evolves with the retrogressive failure in this case.
3.3.1. Slab Failure
As stated above, when the peak shear strength in the sliding layer is high, the failure takes place along the
weak layer, forming a peanut shaped slip surface. However, when the peak shear strength was reduced to
𝐴𝐴 𝐴𝐴uw,p = 𝑠𝑠us,p = 10 kPa, unstable slip surface growth resulted in a slab failure. Figure 7 shows the contours of the
plastic strain and von Mises stress in the overlying layer, together with the continuing slip surface growth in the
weak layer during slab failure. The von Mises stress (see its expression in Appendix A) is often used to deter-
mine if a given material experiences plastic flow. 𝐴𝐴 At 𝐴𝐴 = 5 s, slab failure emerges at the rear of the slip surface
where soils are unloaded and the stress reaches the maximum 20 kPa; while in front of the slip surface, soils are
loaded and the stress decreases from the initial value.
𝐴𝐴 At 𝐴𝐴 = 10 s, the soils in front of the slip surface have been
loaded to the (passive) failure state, and, accordingly, the stress reaches the maximum. Thereafter, the slab failure
propagates mainly at the downslope portion with diffusive plastic strain; at the rear of the slip surface, the plastic
strains are accumulated, but the propagation of the slab failure is not apparent. Instead of the fan zone formed
in front of the slip surface, the rear boundary of the slip surface (or the main scarp) is quite straight to slightly
arcuate. For the planar slope studied here, the slip surface keeps growing along the x-direction, with the side
boundaries of the slip surface continuing to extend further outward.

Figure 8. Investigation of post-failure mass movement in terms of the evolutions of contours of shear strength in weak layer (top row), plastic shear strain in sliding
layer (middle row) and normalized sliding layer thickness (bottom row). Slab failure initiates at upper slope followed by formation of an intact detached block and
disintegration with traveling; mass transport deposits (MTDs) are found at lower basin forming a fan.
3.4. Post-Failure Evolution and Redeposition of Slide Mass
It has been demonstrated above that, once initiated, the slip surface growth and slab failure extension cannot be
arrested on a planar slope. Either slope flattening or material strengthening can restrict slip surface growth and
slab failure extension. To further observe the arrest of slip and post-failure behavior, the full model including
flat basin floors at the margins of the submarine slope, as shown in Figure 2, is used. It consists of a submarine
slope with a slope angle of 6° and length of 400 m (coordinate 𝐴𝐴 𝐴𝐴 from −200 to 200 m), an upper basin floor
𝐴𝐴 (𝐴𝐴 < −200 m) and a lower basin 𝐴𝐴 floor (𝐴𝐴 𝐴 200 m). The peak shear strength of the weak layer soil, the weak layer
depth and the submerged soil unit weight were chosen 𝐴𝐴 as 𝐴𝐴uw,p = 15 𝐴𝐴 kPa, 𝐴 = 8 m,
𝐴𝐴 and 𝐴𝐴 ′ = 8 kN/m 3, respectively.
This parameter set generates a strength ratio 𝐴𝐴 of 𝐴𝐴uw,p ∕𝛾𝛾 ℎ = 0.234, which is typical for normally consolidated
′
marine sediments. The overlying layer was assumed slightly over-consolidated with a peak strength 𝐴𝐴 of 𝐴𝐴us,p = 10
kPa and strength ratio of 0.31. The soil sensitivities of the weak layer and the overlying layer were set to 7 and
2, respectively. An elliptical initial slip surface
𝐴𝐴 of 𝐴𝐴𝑥𝑥 = 40 𝐴𝐴
m and 𝐴𝐴𝑦𝑦 = 80 m was pre-set. Other parameters are the
same as those listed in Table 1.
Figure 8 shows the landslide dynamic evolution from the slab failure initiation, post-failure stage to re-deposition
with respect to the contours of the shear strength in the weak layer and the plastic strain in the overlying layer, and
the changes in the overlying layer thickness. Note that the normalized sliding layer thickness is calculated as the
ratio of the current thickness and the initial thickness of the overlying layer.
𝐴𝐴 At 𝐴𝐴 = 10 s, slab failure is initiated
with a certain amount of slip surface growth. The failure is more concentrated at the rear of the slip surface, and

a curved back scarp is formed at this stage, with the height of the scarp exceeding 4 𝐴𝐴 m (0.5𝐴). The main scarp
moves backward with the retrogressive failure at the rear of the slip surface. The retrogression is stopped by the
flat upper basin floor, leaving a straight main scarp of 200 m in length and over 4 m in height. Because of the
material degradation, the failed overlying mass is torn apart into blocks, as shown in the second row of Figure 8.
The blocks are broken into smaller pieces and finally disappear during their downward movement. The failed and
softened slide mass is finally deposited at the lower basin floor, forming a compressed fan zone of around 400 m
in diameter and over 4 m in thickness. The extent of the slip surface in the weak layer is almost identical to the
combined area of the source region and deposition fan zone.
4. Discussion
4.1. Mechanism and Evolution of Translational Submarine Landslides
Observed from the numerical modeling, the evolution of translational submarine landslides can be summarized
as follows.
1. S lip surface initiation. A slip surface might be formed and grow stably within a weak layer due to external
triggers such as earthquakes and excess pore pressure accumulation. Strain softening of sensitive soils during
shearing leads to the formation of a process zone, where shear strength reduces from the peak toward the
residual, surrounding the slip surface.
2. Slip surface growth. Once the slip surface reaches a certain size, its growth becomes unstable and cata-
strophic, restricted by slope flattening or slab failure only. The critical area of the slip surface for unstable
growth is almost independent of its shape but depends on the material properties and shear stress at the slip
surface. Regardless of the initial shape, the slip surface transitions from an ellipse to a “peanut” pattern during
the unstable growth stage.
3. Slab failure. With the growth of the slip surface, the driving force increases. Therefore, at a certain stage of the
unstable slip surface growth, the overlying soils may reach the maximum allowable stress, initiating slab fail-
ure. After the slab failure, the growth of the slip surface is in alignment with the propagation of the slab failure.
4. Post-failure evolution. A main scarp forms at the rear of the slip surface after the slab failure, and is followed
by retrogression, which is limited by slope flattening. The failed slide mass disintegrates into blocks and then
turns to fully softened debris flow with downward movement. The slide mass finally re-deposits at the lower
basin with the MTDs forming a fan zone.
4.2. Features of Dynamic Slip Surface Evolution

4.2.1. Slip Surface Growth Pattern
The pattern of stable growth of the slip surface depends on the shape of the initial slip surface. Figure 9 shows
the different growth patterns for elliptical slip surfaces with different ratios of major and minor axes, in terms
of the shear strength contour. In all cases, the stable growth of the slip surface together with the development of
the process zone is obvious. As demonstrated above, for a circular slip surface𝐴𝐴with 𝐴𝐴𝑥𝑥 ∕𝑙𝑙𝑦𝑦 = 1, the travel direc-
tion (x-direction) is the favored direction for slip surface growth. This is enhanced with a larger axis ratio (a
wider slip surface), for example,
𝐴𝐴 𝐴𝐴𝑥𝑥 ∕𝑙𝑙𝑦𝑦 = 2. For a slender slip surface (of small axis𝐴𝐴ratio, 𝐴𝐴𝑥𝑥 ∕𝑙𝑙𝑦𝑦 = 1∕3), however,
the slip surface tends to propagate along the y-direction first, presenting a different growth pattern from wide
slip surfaces. For instance,𝐴𝐴with 𝐴𝐴𝑥𝑥 = 20 m 𝐴𝐴 and 𝐴𝐴𝑦𝑦 = 60 m, soil at the two sides of the slip surface begins to fail
while the soil in front and at the rear of the slip surface remains intact. With the axis𝐴𝐴ratio 𝐴𝐴𝑥𝑥 ∕𝑙𝑙𝑦𝑦 = 1∕2, the slip
surface seems to grow simultaneously along the periphery of the slip surface without an obvious favored direc-
tion, which forms the most pessimistic situation.
The growth of the slip surface along the x-direction is driven by the compression force downslope and extension
force upslope, akin to the in-plane shear mode of crack propagation in fracture mechanics (i.e., a shear stress
acting parallel to the plane of the slip surface and perpendicular to the slip surface front); while the growth
along the y-direction is driven by shearing, akin to the out-of-plane shear mode of crack propagation in fracture
mechanics (i.e., a shear stress acting parallel to the plane of the slip surface and parallel to the slip surface front).
Here, the former is defined as the compression-extension mode and the latter is defined as the shear mode. A
combined mode including both the compression-extension and shear modes is expected to be more common,

Figure 9. Three modes of initial slip surface growth: the shear mode (top row) with lateral expansion favored; the compression—extension mode (bottom row) with
upslope and downslope expansion; and the combined mode (middle row) with expansion in all directions.
particularly when the slip surface growth is unstable and continuous. The mechanisms for different growth modes
have been conceptually illustrated in Figure 9. It has not been possible to study the shear and combined modes in
previous 2D investigations (e.g., Kvalstad et al., 2005; Puzrin et al., 2004; Zhang et al., 2015).
Zhang et al. (2020) assumed that the (horizontal) velocity of the slide mass along the x- direction is negligible
compared to the y-direction component, which generates a plane strain condition with the compression/extension
modulus in the sliding layer calculated by
𝐸𝐸 2𝐺𝐺
(10)
𝐸𝐸𝑝𝑝𝑝𝑝 = =
1 − 𝜈𝜈 2 1 − 𝜈𝜈
where 𝐴𝐴 is the Young's modulus

𝐴𝐴 𝐴𝐴 and 𝐴𝐴 the Poisson's ratio. Though this assumption has proved to be not physically
true for the stage of unstable slip surface growth as shown in Figure 6b, it would be helpful to investigate its effect
on slip surface growth and hence determine if it is sufficient for analytical analysis. To achieve this, conserva-
tion of momentum in the x-direction, that is, governing Equation 5, was ignored and the horizontal component
of the velocity was set to zero. Numerical results of such an idealized case in terms of the strength contours are
presented in Figure 6d.𝐴𝐴 For 𝐴𝐴 = 5 s and 10 s, the slip surfaces of the idealized case are almost identical to the case
formulated by rigorous governing equations as shown in Figure 6a. However, with further unstable growth of the
slip surface, the shape of the slip surface remains an ellipse, as opposed to the “peanut” shape of the rigorous case.
Figure 6e compares the two mechanisms 𝐴𝐴 at 𝐴𝐴 = 50 s. The major and minor axes of the “peanut” slip surface are
the same as those of the “ellipse” slip surface, with the area of the slip surface larger in the former mechanism.

The different mechanisms found in the two cases imply that the horizontal movement of the slide mass plays an
important role and must be considered during the unstable growth of the slip surface.
4.2.2. Slip Surface Growth Speed
Once a slip surface falls into the unstable growth stage, the growth speed depends on how the unbalanced forces
are transferred within the overlying layer. For the compression-extension mode, the growth of the slip surface is
driven by the compressional/tensile force, and therefore the growth speed (of the major axis) can be related to the
compression wave velocity
√
𝐸𝐸 ′
(11) 𝑣𝑣𝑚𝑚𝑚𝑚𝑚𝑚 = 2
𝜌𝜌
where 𝐴𝐴 ′ is the compression modulus. Note that the number 2 in the expression means that the growth speed is
𝐴𝐴
double the wave velocity, as the slip surface grows in both the upslope and downslope directions. Similarly, for
the shear mode, the growth speed of the minor axis can be related to the shear wave velocity
√
𝐺𝐺
(12) 𝑣𝑣𝑚𝑚𝑚𝑚𝑚𝑚 = 2
𝜌𝜌
For plane strain and undrained conditions, the compression modulus,

𝐴𝐴 𝐴𝐴 ′ = 𝐸𝐸ps, is four times the shear modulus
(see Equation 10) and therefore, the major axis always doubles the minor axis of the slip surface. This can be
seen in Figure 6e, where the major axes of the slip surfaces are almost 3,000 m while the minor axes are around
1,500 m in both mechanisms.
Figure 10a shows the length (major axis) and width (minor axis) of the slip surface during its growth for the
selected case, compared with the analytical solutions given by Equations 11 and 12. It should be noted that the
growth of the slip surface evolves from the stable to unstable stages with the transition emerging at around
𝐴𝐴 𝐴𝐴 =
4 s. During the stable growth stage, both axes of the slip surface are assumed unchanged. With this idealization,
the growth of the two axes of the slip surface in the numerical modeling can be well predicted by the analytical
solutions, with the growth speeds being 34 m/s and 68 m/s for the minor and major axes, respectively. Such fast
speeds reveal that unstable growth of the slip surface is catastrophic and significantly differs from creep failure.
Figure 10b gives the area of the slip surface during its unstable growth for both mechanisms. For the “ellipse”
mechanism, the area can be calculated exactly by
𝜋𝜋
(13)
𝐴𝐴 = 𝑙𝑙𝑥𝑥 𝑙𝑙𝑦𝑦
4
which is shown by the good agreement between the numerical and analytical results in the figure. The area of
the ellipse slip surface is initially 1,962.5 m 2 and increases to 3.6 km 2 in 50 s. The peanut slip surface (5.3 km 2
calculated from the numerical modeling results) is about 45% larger than the ellipse slip surface𝐴𝐴 at 𝐴𝐴 = 50 s. The
fast growth of the slip surface implies that during an earthquake, even with a short period of shaking, a large slip
surface with a magnitude of several km 2 might be formed. Such a large slip surface may further result in slab
failure and debris flow. Therefore, it is key to determine in what conditions the slip surface can grow unstably,
which will be discussed in the next sub-section.
4.3. Criteria for Unstable Growth of Slip Surface
For a slip surface described by a series of functions Equation 9, the area of the slip surface can be calculated by
𝑙𝑙𝑥𝑥 𝑙𝑙𝑦𝑦 Γ(1 + 1∕𝑛𝑛)Γ(1∕𝑛𝑛)

(14)
𝐴𝐴 = ⋅
𝑛𝑛 Γ(1 + 2∕𝑛𝑛)
where
𝐴𝐴 Γ is the gamma function. By integrating the normal and shear resistances along the boundary of the slip
surface, one may calculate the total resistance and compare it to the driving force from the slip surface, whereby
the critical area of the slip surface for unstable growth is given by Zhang et al. (2020)
( )
1 − 𝑟𝑟 2
(15)
𝐴𝐴𝑐𝑐𝑐𝑐𝑐𝑐 = 32 𝑙𝑙𝑐𝑐
𝑟𝑟

Figure 10. Speed of slip surface growth in terms of (a) lengths of major and minor axes and (b) area of slip surface; (c) critical area of slip surface for unstable growth
by numerical and analytical analyses; and (d) best fitting of numerical data.
where 𝐴𝐴 is the shear stress ratio

𝐴𝐴 𝐴𝐴 and 𝐴𝐴c is the characteristic length relevant to the process zone size, given by
√
𝜏𝜏𝑔𝑔 − 𝑠𝑠𝑢𝑢𝑢𝑢𝑢𝑢𝑢 𝐺𝐺𝐺𝐺𝐺𝑟𝑟𝑝𝑝
(16) 𝑟𝑟 = , 𝑙𝑙𝑐𝑐 =
𝑠𝑠𝑢𝑢𝑢𝑢𝑢𝑢𝑢 − 𝑠𝑠𝑢𝑢𝑢𝑢𝑢𝑢𝑢 𝑠𝑠𝑢𝑢𝑢𝑢𝑢𝑢𝑢 − 𝑠𝑠𝑢𝑢𝑢𝑢𝑢𝑢𝑢
For static analysis, ignoring any inertia effects, the Equation 15 is conservative compared to the numerical data
from finite element and finite difference modeling (Zhang et al., 2020). However, this should be verified under
dynamic conditions which can be achieved using the proposed numerical modeling.
A parametric study was conducted to observe the effects of the shape parameter, 𝐴𝐴 𝐴𝐴, and the dimensions of the
slip surface on the critical area for unstable slip surface growth. The gravity loads and the critical surface areas at
critical conditions for all cases are presented in Table 2. Figure 10c shows a comparison of the numerical results
with or without inertia effects and the analytical results by Equation 15. With inertia effects, the critical area
estimated by Equation 15 is not always conservative, particularly with a large shear stress ratio. If the dynamic
criterion meets the same series of functions as the static Equation 15, the best fit of the numerical data from the
dynamic analysis gives
( )
1 − 𝑟𝑟 2
(17)
𝐴𝐴𝑐𝑐𝑐𝑐𝑐𝑐 = 18.4 𝑙𝑙𝑐𝑐
𝑟𝑟
which is shown in Figure 10d. This means that the critical area of the slip surface for unstable slip surface growth
under dynamic conditions is on average 42.5% smaller than that ignoring inertia effects. This echoes the finding
by Zhang et al. (2016) that for a 2D cross-section of a planar slope, the critical length (major axis) of slip surface
for catastrophic propagation can be up to 50% lower with inertia effects than without inertia effects. Extending

Table 2
Critical Conditions for Unstable Slip Surface Growth by Numerical Analysis
Dimension of slip surface Gravity load
𝐴𝐴 𝒍𝒍𝒙𝒙 (m) 𝐴𝐴 𝒍𝒍𝒚𝒚 (m) 𝐴𝐴 𝒏𝒏 𝐴𝐴 𝑨𝑨 (m )2

𝐴𝐴 𝝆𝝆 (kg/m )3
𝐴𝐴 𝒓𝒓
First 𝐴𝐴
series: 𝐴𝐴𝑦𝑦 ∕𝑙𝑙𝐴𝐴𝑥𝑥 = 1, 𝐴𝐴 = 2
10 10 2 79 1,170 0.943
20 20 2 314 1,070 0.841
40 40 2 1,257 870 0.637
60 60 2 2,827 730 0.494
80 80 2 5,027 640 0.403
100 100 2 7,854 580 0.341
120 120 2 11,310 540 0.301
140 140 2 15,394 500 0.260
160 160 2 20,106 470 0.229
180 180 2 25,447 450 0.209
200 200 2 31,416 430 0.189
Second 𝐴𝐴
10 20 2 157 1,130 0.902
20 40 2 628 980 0.749
40 80 2 2,513 750 0.515
60 120 2 5,655 630 0.392
80 160 2 10,053 550 0.311
100 200 2 15,708 500 0.260
120 240 2 22,619 460 0.219
140 280 2 30,788 430 0.189
Third 𝐴𝐴
10 30 2 236 1,090 0.862
20 60 2 942 910 0.678
40 120 2 3,770 680 0.443
60 180 2 8,482 570 0.331
80 240 2 15,080 500 0.260
100 300 2 23,562 460 0.219
120 360 2 33,929 430 0.189
Fourth 𝐴𝐴
series: 𝐴𝐴𝑦𝑦 ∕𝑙𝑙𝑥𝑥 𝐴𝐴= 0.5, 𝐴𝐴 = 2
20 10 2 157 1,130 0.902
40 20 2 628 990 0.760
80 40 2 2,513 770 0.535
120 60 2 5,655 640 0.403
160 80 2 10,053 560 0.321
200 100 2 15,708 510 0.270
240 120 2 22,619 470 0.229
280 140 2 30,788 450 0.209
Fifth 𝐴𝐴
20 40 1 400 1,040 0.811
40 80 1 1,600 830 0.596
60 120 1 3,600 700 0.464

Table 2
Continued
Dimension of slip surface Gravity load
𝐴𝐴 𝒍𝒍𝒙𝒙 (m) 𝐴𝐴 𝒍𝒍𝒚𝒚 (m) 𝐴𝐴 𝒏𝒏 𝐴𝐴 𝑨𝑨 (m )

2
𝐴𝐴 𝝆𝝆 (kg/m )
3
𝐴𝐴 𝒓𝒓
80 160 1 6,400 610 0.372
100 200 1 10,000 550 0.311
120 240 1 14,400 510 0.270
Sixth 𝐴𝐴
20 40 10 789 950 0.719
40 80 10 3,154 720 0.484
60 120 10 7,097 590 0.352
80 160 10 12,617 520 0.280
100 200 10 19,715 470 0.229
120 240 10 28,389 440 0.199
this observation to the 3D case, one may simply assume for the dynamic unstable growth a 50% reduction in the
critical area of the slip surface from the static Equation 17, that is
( )
1 − 𝑟𝑟 2
(18)
𝐴𝐴𝑐𝑐𝑐𝑐𝑐𝑐 = 16 𝑙𝑙𝑐𝑐
𝑟𝑟
Figure 10c shows that Equation 18 gives estimates of the critical slip surface area well below the numerical data
and is therefore conservative.
4.4. Effects of a 3D Slope Geometry
Three typical 3D slope types, as shown in Figures 2b–2d, are considered in this section in order to gain an
initial insight into the slope geometry effects on the translational landslide evolution. For the S-shape slope, the
half-height of the slope in Equation 1 was set to be the same as for the planar slope, that 𝐴𝐴 is, 𝐴𝐴 = 21 m, and the
maximum slope angle was taken 𝐴𝐴 as 𝐴𝐴𝑐𝑐 = 9◦ such that the average slope angle within the range of −500 m < 𝑦 < 500
m is equal to the planar slope angle of 6°. For the convex slope, the values 𝐴𝐴 of 𝐴𝐴𝐴𝐴𝑡𝑡 and 𝐴𝐴𝑏𝑏 were set to 800 m and
1,200 m, respectively; while for the concave 𝐴𝐴 slope, 𝐴𝐴𝑏𝑏 = 800 𝐴𝐴
m and 𝐴𝐴𝑡𝑡 = 1, 200 m. The slope angle of the convex
and concave slope models is the same with the planar slope model.
Figure 11 compares the final states of the four slope models with respect to the fields of the shear strength in the
weak layer, the shear stress in the weak layer, the plastic strain in the sliding layer, and the normalized sliding
layer thickness. It can be noted that the far field gravity shear stress fields (second row of Figure 11) strongly
depend on the geometry of the problem. The results of the planar, convex and concave slopes look very similar,
with slightly more horizontal slip surface growth observed in the convex slope and slightly more retrogressive
extension pertained in the concave slope, suggesting that the slope gradient along the x-direction has limited
influence on the landslide evolution. In contrast, the final slip surface and MTD observed in the S-shape curvi-
linear slope are significantly different from the other three models with less extended retrogressive failure and a
smaller fan heave zone. This indicates that the slope gradient along the y-direction has a considerable effect on
the landslide evolution.
The examples shown here present an initial investigation of 3D post-failure behavior of submarine landslides.
Detailed investigation of the effects of the 3D slope geometry on post-failure patterns is beyond the scope of the
present work and will be explored in future studies by using the proposed numerical tool.

Figure 11. 3D slope geometry effects on the ultimate slip surface growth and morphology of the mass transport deposit
(MTD) in submarine landslides in terms of: slope geometry (first row); shear strength in weak layer (second row); shear
stress in weak layer (third row); plastic shear strain in sliding layer (forth row); and (d) normalized sliding layer thickness
(fifth row). Note: four types of slope geometries—planar (first column), S-shape (second column), convex (third column)
and concave (forth column)—are studied with details given in Figure 2; shear strength in weak layer reflects slip surface
configuration; plastic strain in sliding layer reflects the intactness of slide mass; normalized thickness is calculated as current
thickness over initial thickness.
5. Conclusions
Submarine landslides are a major marine geohazard posing significant threats to offshore infrastructure and coastal
urban centers. This study has provided insights into the entire evolution of submarine landslides in three dimen-
sional (3D) slopes by using an original large deformation computational tool. For the first time, the paper combines
within a single framework the modeling of both pre-failure initiation and post-failure evolution of 3D submarine
landslides. This facilitates understanding of the physical mechanisms behind the development of an enormous
submarine landslide from a small slip surface and helps to quantify different stages of the failure evolution.

The findings from the study may advance scientific understanding of historic events and help safeguarding
offshore developments against submarine landslide risks.
The complete evolution of submarine translational landslides includes the slip surface initiation and growth along
a weak layer, slab failure, post-failure mass movement and re-stabilization, as depicted in Figure 1. A slip surface
might be formed with a relatively small size and grow in a stable manner due to external triggers such as earth-
quakes and excess pore pressure accumulation. The growth pattern depends on the shape of initial slip surface.
For an initially wide slip surface, the process zone first emerges in front and at the rear of the slip surface, whereas
for an initially slender slip surface, it occurs at the two sides. For an elliptical slip surface, the process zone devel-
ops around the periphery of the slip surface without any favored direction. The critical area of the slip surface for
unstable growth is almost independent of its shape but depends on the material properties and shear stress ratio
( )2
over the slip surface. For planar submarine slopes, it is given 𝐴𝐴 by 𝐴𝐴cri = 16 1 −𝑟𝑟 𝑟𝑟 𝑙𝑙c where the shear stress ratio,
𝐴𝐴 𝐴𝐴, and characteristic length,
𝐴𝐴 𝐴𝐴c, are expressed by Equation 16.
Regardless of the initial shape, the slip surface transitions from an ellipse to a “peanut” pattern during the unstable
growth stage, with expansion rates equal to compression wave velocity and shear wave velocity along the major
and minor axes of the slip surface, respectively. The slab failure usually initiates at the rear of the slip surface if
the horizontal earth pressure is smaller than the vertical one, which is usually the case in submarine sediments.
The stronger the overlying layer, the larger the slip surface before the slab failure. A main scarp forms at the rear
of the slip surface after the slab failure, and is followed by retrogression, which is limited by upslope slope flat-
tening. Once slab failure is triggered, slip surface growth is controlled by upslope retrogression and downslope
plowing, and hence its shape deviates from a “peanut.” The slide mass finally re-deposits at the flat terrain with
the MTD forming a fan zone. The differences in the landslide failure extension and MTD morphology between
the planar, the convex, and the concave slopes of the same and uniform slope angle and parallel layering char-
acteristic of sediments were found to be insignificant. In contrast, in the curvilinear slope, a significantly less
extended failure upslope and a smaller fan heave zone downslope have been observed.
Appendix A: Numerical Scheme for 3D Modeling of Submarine Landslide Evolution

The domain of interest, that covers the source and MTD areas of a submarine landslide, is essentially divided
into regularized cells, with each cell holding characteristics of the evolving landslide. The edges of the cell are
parallel to the axes of coordinates x and y, and the x–y plane (z = 0) was set as the horizontal plane and crossing
through a reference point (taken as the slope center in the study) at the basal slip surface. Cells are fixed during
the landslide process, with materials traveling through them, forming a Eulerian framework. Conservations of
mass and momentum are then formulated within each cell, and global instability can be modeled by integrating
all cells with consideration of proper inter-cell constitutive models and fluxes.
Conservation of mass in each cell can be expressed by
𝜕𝜕𝜕 𝜕𝜕𝜕𝜕𝜕 𝜕𝜕𝜕𝜕𝜕

(A1) + + =0
𝜕𝜕𝜕𝜕 𝜕𝜕𝜕𝜕 𝜕𝜕𝜕𝜕
where 𝐴 is the thickness of the slab above the weak𝐴𝐴layer,𝐴𝐴𝐴𝐴 and 𝐴𝐴 are the velocity𝐴𝐴in the 𝐴𝐴
𝐴𝐴 𝐴𝐴- and 𝐴𝐴-directions, respec-
tively,
𝐴𝐴 and 𝐴𝐴 is the elapsed time. Conservation of momentum in each cell is given by
𝜕𝜕𝜕𝜕𝜕 𝜕𝜕𝜕𝜕𝜕2 𝜕𝜕𝜕𝜕𝜕𝑥𝑥 𝜕𝜕𝜕𝜕𝜕𝜕𝜕 𝜕𝜕𝜕𝜕𝜕𝑥𝑥𝑥𝑥 𝜏𝜏𝑤𝑤𝑤𝑤𝑤 + 𝜏𝜏𝑔𝑔𝑔𝑔𝑔 + 𝜏𝜏𝑑𝑑𝑑𝑑𝑑𝑑𝑑𝑑𝑑𝑑𝑑

(A2) + + + − − =0
and
𝜕𝜕𝜕𝜕𝜕 𝜕𝜕𝜕𝜕𝜕2 𝜕𝜕𝜕𝜕𝜕𝑦𝑦 𝜕𝜕𝜕𝜕𝜕𝜕𝜕 𝜕𝜕𝜕𝜕𝜕𝑥𝑥𝑥𝑥 𝜏𝜏𝑤𝑤𝑤𝑤𝑤 + 𝜏𝜏𝑔𝑔𝑔𝑔𝑔 + 𝜏𝜏𝑑𝑑𝑑𝑑𝑑𝑑𝑑𝑑𝑑𝑑𝑑

(A3) + + + − − =0
for
𝐴𝐴 the 𝐴𝐴𝐴𝐴- and 𝐴𝐴-directions, respectively. In the above equations,
𝐴𝐴 𝐴𝐴 is the bulk density of saturated 𝐴𝐴 soil,
𝐴𝐴 𝐴𝐴𝑥𝑥, 𝐴𝐴𝑦𝑦 and
𝐴𝐴 𝐴𝐴𝑥𝑥𝑥𝑥 are stress components applied at the center of the cell face, with the face normals parallel 𝐴𝐴 to the
𝐴𝐴 𝐴𝐴 or 𝐴𝐴 axis;
𝐴𝐴 𝐴𝐴 and 𝐴𝐴w,𝑦𝑦 are weak layer (or slip surface) shear stress components;
𝐴𝐴w,𝑥𝑥 𝐴𝐴 𝐴𝐴 and 𝐴𝐴g,𝑦𝑦 are shear stress components,
𝐴𝐴g,𝑥𝑥

associated with the gravity, at the buried depth of the weak layer; 𝐴𝐴 and 𝐴𝐴drag,𝑦𝑦 are shear stress components
𝐴𝐴 and 𝐴𝐴drag,𝑥𝑥
associated with the drag of ambient water.
A1. Stress Components at Cell Face Center

Usually, the stress tensor
𝐴𝐴 (𝝈𝝈 ) describing the stress status at the center of the cell face can be decomposed into
(A4)
𝝈𝝈 = 𝒔𝒔 + 𝑝𝑝 ⋅ 𝑰𝑰
where
𝐴𝐴 𝒔𝒔 is the deviatoric stress tensor,
𝐴𝐴 𝐴𝐴 the mean stress
𝐴𝐴 and 𝑰𝑰 the second-order identity tensor. Note that the
vertical normal stress component
𝐴𝐴 𝐴𝐴𝑧𝑧 can be expressed by
1
(A5)
𝜎𝜎𝑧𝑧 = 𝛾𝛾 ′ ℎ
2
where 𝐴𝐴 ′ = 𝜌𝜌′ 𝑔𝑔 is the submerged unit weight of soils𝐴𝐴with 𝐴𝐴′ being the submerged density.
𝐴𝐴
Similarly, the strain tensor

𝐴𝐴 (𝜺𝜺) can be decomposed into
𝜀𝜀
𝜺𝜺 = 𝒆𝒆 + 𝑣𝑣 ⋅ 𝑰𝑰
(A6)
3
where 𝒆𝒆 is the deviatoric strain tensor

𝐴𝐴 𝐴𝐴 and 𝐴𝐴v the volumetric strain. It can also be divided into the elastic and plastic
portions, which will be denoted by superscripts ‘e’ and ‘p’, respectively, in the remainder of the paper. Note that
the volumetric strain satisfies
𝐴𝐴 𝐴𝐴v = 𝜀𝜀ev = 𝜀𝜀v → 0, as the undrained condition was maintained and the von Mises
p
yield criterion with an associated flow rule was used. Hence, one may𝐴𝐴write 𝜺𝜺𝐴𝐴≅ 𝒆𝒆, 𝜺𝜺e ≅ 𝒆𝒆𝐴𝐴 e
, and 𝜺𝜺p ≅ 𝒆𝒆p. The
elasticity of materials is assumed linear and isotropic, and therefore the deviatoric stress tensor is expressed by
(A7)
𝒔𝒔 = 2𝐺𝐺𝒆𝒆𝑒𝑒 ≅ 2𝐺𝐺𝜺𝜺𝑒𝑒
where 𝐴𝐴 is the shear modulus.

𝐴𝐴
A modified von Mises yield criterion was adopted in order to consider isotropic and linear strain softening,
given by
( )
𝑝
1
𝑞 = 𝑚𝑎𝑥 1 − 𝑝 ,
(A8)
𝜀𝑠
⋅ 2𝑠𝑢𝑠,𝑝
𝜀𝑠,𝑟 𝑆𝑡
√
where 𝐴𝐴 =
𝐴𝐴 3
‖𝒔𝒔‖ is
the von Mises 𝐴𝐴
stress, 𝐴𝐴us,p the peak undrained shear strength in the sliding layer which can
2 √
𝐴𝐴 test; 𝐴𝐴s = ∫0 23 ‖𝒆𝒆̇ p ‖𝑑𝑑𝑑𝑑 the accumulated von Mises 𝐴𝐴
be measured from a triaxial element strain; 𝐴𝐴ps,r the value
𝐴𝐴 of 𝐴𝐴s
p 𝑡𝑡 p
to the residual shear strength;

𝐴𝐴 𝐴𝐴t the soil sensitivity defining the ratio of the peak and residual shear strengths.
𝐴𝐴
p
𝐴𝐴s,r
can be determined from a triaxial test𝐴𝐴 by 𝐴𝐴s,r = 3 𝛾𝛾r 𝐴𝐴
p 2 p
where 𝐴𝐴r is the plastic shear strain associated to the residual
p
undrained shear strength.
A2. Shear Stress at Weak Layer

Within the slip surface, the shear stress
𝐴𝐴 (𝐴𝐴w) is limited to the current shear strength, which is reduced during
shearing, and given by
( )
𝑝
𝜏𝑤 = 𝑠𝑢𝑤 (𝛿 𝑝 ) = 𝑚𝑎𝑥 1 − 𝛿𝑝 , 1 ⋅ 𝑠𝑢𝑤,𝑝
(A9)
𝛿 𝑆 𝑟 𝑡
where 𝐴𝐴 p = ∫0 ‖𝜹𝜹̇ ‖𝑑𝑑𝑑𝑑 is the accumulated plastic shear displacement across the weak𝐴𝐴layer, 𝐴𝐴rp the value 𝐴𝐴 of 𝐴𝐴 p at the
𝑡𝑡 p
𝐴𝐴
residual shear stress, 𝐴𝐴 and 𝐴𝐴uw,p the peak undrained shear strength in the weak layer. Ignoring displacement beneath
the weak layer (Zhang et al., 2015), the horizontal slide displacement can be related to the shear displacement
across the weak layer 𝐴𝐴 by cos𝑢𝑢 𝜃𝜃 = 𝛿𝛿 = 𝛿𝛿 e + 𝛿𝛿 p 𝐴𝐴
where 𝐴𝐴 is the slope angle
𝐴𝐴 and 𝐴𝐴
𝐴𝐴 e and 𝐴𝐴 p the elastic and plastic portion
of the shear displacement, respectively.

Soils surrounding the slip surface are first mobilized elastically before reaching the yield stress governed by
Equation A9, and the shear stress is increased to be larger than the initial value caused by gravity. Considering a
linear and isotropic elasticity model, the pre-peak shear stress can be expressed by
(A10)
𝝉𝝉 𝑤𝑤 = 𝐾𝐾𝜹𝜹𝑒𝑒
where 𝐴𝐴 is the shear stiffness.

𝐴𝐴
A3. Finite Volume Scheme
Two layers of fixed meshes with the same mesh size and alignment were taken, as shown in Figure A1a, with the
top layer used for solving mass and momentum conservation equations and the bottom layer tracking the changes
in soil properties in the weak layer during slip surface growth. A finite volume method with staggered grids, as
shown in Figure A1b, was used to integrate and solve the governing Equations A1 to A3.
Let us consider a slope of a rectangular space domain 𝐴𝐴 Ω ∶ (0, 𝐿𝐿𝑥𝑥 ) × (0, 𝐿𝐿𝑦𝑦 ) and a time interval
𝐴𝐴 (0, 𝑇𝑇 ). Dirichlet
boundary conditions, that 𝐴𝐴 is, 𝐴𝐴 =𝐴𝐴0 and 𝐴𝐴 = 0, are prescribed representing unaffected remote regions. The space
domain is meshed with a grid 𝐴𝐴 of 𝐴𝐴𝑥𝑥 × 𝑁𝑁𝑦𝑦 cells, and the cells of dimensions 𝐴𝐴 ∆𝑥𝑥𝐴𝐴 and ∆𝑦𝑦 are indexed𝐴𝐴 by (𝑖𝑖𝑖 𝑖𝑖) where
𝐴𝐴 𝐴𝐴 ∈ (0, 𝑁𝑁𝐴𝐴 𝑥𝑥 ) and 𝐴𝐴 ∈ (0, 𝑁𝑁𝑦𝑦 ). The centers of the bottom, top, left, and right edges of the 𝐴𝐴 cell (𝑖𝑖𝑖 𝑖𝑖) are denoted by
( )( )( ) ( )
𝐴𝐴 𝑖𝑖𝑖 𝑖𝑖𝐴𝐴− 21 , 𝑖𝑖𝑖 𝑖𝑖𝐴𝐴+ 12 , 𝑖𝑖 − 12 , 𝑗𝑗𝐴𝐴 , and 𝑖𝑖 + 21 , 𝑗𝑗 , respectively. The mass conservation is integrated and solved over
the cell, with the thickness of the sliding𝐴𝐴layer, 𝐴, and slope angle (topography), 𝐴𝐴 𝐴𝐴, discretized at the cell center.
The velocity in the x-direction is discretized at the center of the edges normal to the x-direction, while the velocity
in the y-direction is discretized at the center of the edges normal to the y-direction. ( ) The approximation 𝐴𝐴 of 𝐴 at cell
𝐴𝐴 (𝑖𝑖𝑖 𝑖𝑖) and𝐴𝐴time 𝐴𝐴𝑛𝑛 is denoted
𝐴𝐴 by 𝐴𝑛𝑛𝑖𝑖𝑖𝑖𝑖. The approximation
( ) 𝐴𝐴 of 𝐴𝐴 at the
𝐴𝐴 edge 𝑖𝑖 + 1
2
, 𝑗𝑗 and𝐴𝐴time 𝐴𝐴𝑛𝑛
is denoted
𝐴𝐴 by 𝐴𝐴𝑛𝑛 1
𝑖𝑖+ 2 ,𝑗𝑗
while the approximation 𝐴𝐴 of 𝐴𝐴 at the 𝐴𝐴 edge 𝑖𝑖𝑖 𝑖𝑖 + 2 and𝐴𝐴time 𝐴𝐴 is denoted
1 𝑛𝑛
𝐴𝐴 𝐴𝐴 1 .
𝑛𝑛
𝑖𝑖𝑖𝑖𝑖+ 2
Sliding layer
Depth averaged Weak layer
planes for meshes
(a)
(c)
Mass
Momentum in
the x-direcon
Momentum in
the y-direcon
(b) (d)
Figure A1. (a) Schematics of depth integrated model; (b) staggered mesh scheme; (c) update of properties for the fixed weak layer; and (d) update of properties for the
movable sliding layer.

At𝐴𝐴time 𝐴𝐴𝑛𝑛+1, the mass conservation Equation A1 is discretized as

( ) ( )
∆𝑡𝑡 ∆𝑡𝑡
(A11)
ℎ𝑛𝑛+1
𝑖𝑖𝑖𝑖𝑖 − ℎ 𝑛𝑛
𝑖𝑖𝑖𝑖𝑖 = 𝑞𝑞𝑞𝑞 𝑛𝑛
𝑖𝑖− 21 ,𝑗𝑗
− 𝑞𝑞𝑞𝑞 𝑛𝑛
+ 𝑞𝑞𝑞𝑞 𝑛𝑛
𝑖𝑖− 12 ,𝑗𝑗
− 𝑞𝑞𝑞𝑞 𝑛𝑛
∆𝑥𝑥 ∆𝑦𝑦
where
⎧ ℎ𝑛 𝑢𝑛 ≤0
⎪ 𝑖,𝑗 1
⎪ 𝑖− ,𝑗
2
𝑞𝑥 1 = ℎ̂ 1 𝑢 1 , ℎ̂ 1 = ⎨
𝑛 𝑛 𝑛 𝑛
;
𝑖− 2 ,𝑗 𝑖− 2 ,𝑗 𝑖− 2 ,𝑗 𝑖− 2 ,𝑗
⎪ℎ𝑛𝑖−1,𝑗 𝑢𝑛 >0
⎪ 1
⎩ 𝑖− ,𝑗
2
≤0
(A12)
⎧ ℎ𝑛 𝑣 𝑛
⎪ 𝑖,𝑗 1
⎪ 𝑖,𝑗−
2
𝑞𝑦𝑛 1 = ℎ̂ 𝑛 1 𝑣𝑛 1 , ℎ̂ 𝑛 1 = ⎨
𝑖,𝑗− 2 𝑖,𝑗− 2 𝑖,𝑗− 2 𝑖,𝑗− 2
⎪ℎ𝑛𝑖,𝑗−1 𝑣𝑛 >0
⎪ 1
⎩ 𝑖,𝑗−
2
The momentum conservation in the x-direction, that is, Equation A2, is discretized as
ℎ𝑛+11 𝑢𝑛+11 − ℎ𝑛 𝑢𝑛
𝑖+ 2 ,𝑗 𝑖+ 2 ,𝑗 𝑖+ 12 ,𝑗 𝑖+ 12 ,𝑗
( )
∆𝑡
= 𝑞𝑥𝑛𝑖,𝑗 𝑢̂ 𝑛𝑖,𝑗 − 𝑞𝑥𝑛𝑖+1,𝑗 𝑢̂ 𝑛𝑖+1,𝑗
∆𝑥
( )
∆𝑡
+ 𝑞𝑦 1 1 𝑢̂ 1 1 − 𝑞𝑦 1 1 𝑢̂ 1 1
𝑛 𝑛 𝑛 𝑛
∆𝑦 𝑖+ 2 ,𝑗− 2 𝑖+ 2 ,𝑗− 2 𝑖+ 2 ,𝑗+ 2 𝑖+ 2 ,𝑗+ 2 (A13)
( )
∆𝑡
+ 𝜏 𝑛 1 1 ℎ𝑛 1 1 − 𝜏 𝑛 1 1 ℎ𝑛 1 1
𝜌∆𝑦 𝑥𝑦,𝑖+ 2 ,𝑗+ 2 𝑖+ 2 ,𝑗+ 2 𝑥𝑦,𝑖+ 2 ,𝑗− 2 𝑖+ 2 ,𝑗− 2
𝜏 1 +𝜏 1 +𝜏
𝑛 𝑛 𝑛
𝑤,𝑖+ 2 ,𝑗 𝑔,𝑖+ 2 ,𝑗 𝑑𝑟𝑎𝑔,𝑖+ 12 ,𝑗
+
𝜌
where
ℎ𝑛𝑖,𝑗 + ℎ𝑛𝑖+1,𝑗
ℎ𝑛 = ;
𝑖+ 12 ,𝑗 2
⎧𝑢𝑛 𝑞𝑥𝑛𝑖,𝑗 ≤ 0
𝑞𝑥𝑛 + 𝑞𝑥𝑛 ⎪ 1
𝑖− 12 ,𝑗 𝑖+ 21 ,𝑗 ⎪ 𝑖+ ,𝑗
𝑞𝑥𝑛𝑖,𝑗 = , 𝑢̂ 𝑛𝑖,𝑗 = ⎨ 2 ;
2 ⎪𝑢𝑛 𝑞𝑥𝑛𝑖,𝑗 > 0
⎪ 𝑖− 1 ,𝑗
⎩ 2
(A14)
⎧ 𝑢𝑛 𝑞𝑦𝑛 ≤0
𝑞𝑦𝑛 1 + 𝑞𝑦𝑛 ⎪ 1 1 1
𝑖,𝑗− 2 𝑖+1,𝑗− 12 ⎪ 𝑖+ ,𝑗 𝑖+ ,𝑗−
𝑞𝑦𝑛 1 1 = , 𝑢̂ 𝑛 1 1 = ⎨ 2 2 2 ;
𝑖+ 2 ,𝑗− 2 2 𝑖+ 2 ,𝑗− 2
⎪𝑢𝑛 𝑞𝑦𝑛 >0
⎪ 𝑖+ 1 ,𝑗−1 1 1
⎩ 2 𝑖+ ,𝑗−
2 2
𝜏 𝑛 1 1 ℎ𝑛 1 1
𝑥𝑦,𝑖+ 2 ,𝑗+ 2 𝑖+ 2 ,𝑗+ 2
𝑛
𝜏𝑥𝑦,𝑖,𝑗 ℎ𝑛𝑖,𝑗 + 𝜏𝑥𝑦,𝑖+1,𝑗
𝑛
ℎ𝑛𝑖+1,𝑗 + 𝜏𝑥𝑦,𝑖,𝑗+1
𝑛
ℎ𝑛𝑖,𝑗+1 + 𝜏𝑥𝑦,𝑖+1,𝑗+1
𝑛
ℎ𝑛𝑖+1,𝑗+1
=
4
The momentum conservation in the y-direction, that is, Equation A3, can be discretized in a similar way.

Data Availability Statement

Modeling for this research was conducted using an open-source Python package developed by the first author
and available at https://github.com/geotechzhang/TranSliM. The solver has been validated against a historic event
occurred on the eastern margin of New Zealand (Watson et al., 2020) with satisfying outcomes. Data used in this
study is publicly available at the Figshare repository https://doi.org/10.6084/m9.figshare.17029700.
Acknowledgments References
The authors would like to thank Honorary
Prof. Mark Randolph of the University of Biscarini, C. (2010). Computational fluid dynamics modelling of landslide generated water waves. Landslides, 7(2), 117–124. https://doi.
Western Australia and Prof. Dong Wang org/10.1007/s10346-009-0194-z
of the Ocean University of China for valu- Cornforth, D. H. (2005). Landslides in practice: Investigation, analysis and remedial/preventative options in soils. John Wiley and Sons Inc.
able discussions on the topic. Open access De Blasio, F. V., Engvik, L., Harbitz, C. B., & Elverhøi, A. (2004). Hydroplaning and submarine debris flows. Journal of Geophysical Research,
funding provided by Eidgenossische 109(C1), C01002. https://doi.org/10.1029/2002jc001714
Technische Hochschule Zurich. Dong, Y., Wang, D., & Randolph, M. F. (2017). Runout of submarine landslide simulated with material point method. Procedia Engineering, 175,
357–364. https://doi.org/10.1016/j.proeng.2017.01.045
Elverhøi, A., Issler, D., De Blasio, F. V., Ilstad, T., Harbitz, C. B., & Gauer, P. (2005). Emerging insights into the dynamics of submarine debris
flows. Natural Hazards and Earth System Science, 5, 633–648. https://doi.org/10.5194/nhess-5-633-2005
Frey-Martínez, J., Cartwright, J., & James, D. (2006). Frontally confined versus frontally emergent submarine landslides: A 3D seismic charac-
terisation. Marine and Petroleum Geology, 23(5), 585–604. https://doi.org/10.1016/j.marpetgeo.2006.04.002
Haflidason, H., Sejrup, H. P., Nygard, A., Mienert, J., Bryn, P., Lien, R., et al. (2004). The Storegga slide: Architecture, geometry and slide
development. Marine Geology, 213(1–4), 201–234. https://doi.org/10.1016/j.margeo.2004.10.007
Imran, J., Harff, P., & Parker, G. (2001). A numerical model of submarine debris flow with graphical user interface. Computers & Geosciences,
27(6), 717–729. https://doi.org/10.1016/s0098-3004(00)00124-2
Issler, D., L’Heureux, J. S., Cepeda, J. M., & Quan Luna, B. (2015). Towards a numerical run-out model for quick-clay slides. In EGU general
assembly 2015.
Kim, J., Løvholt, F., Issler, D., & Forsberg, C. F. (2019). Landslide material control on tsunami Genesis—The Storegga slide and tsunami (8, 100
years BP). Journal of Geophysical Research: Oceans, 124(6), 3607–3627. https://doi.org/10.1029/2018jc014893
Klein, B., & Puzrin, A. M. (2021). Growth of slip surfaces in 3D conical slopes. International Journal for Numerical and Analytical Methods in
Geomechanics, 45(12), 1683–1711. https://doi.org/10.1002/nag.3220
Kvalstad, T. J., Andresen, L., Forsberg, C. F., Berg, K., Bryn, P., & Wangen, M. (2005). The Storegga slide: Evaluation of triggering sources and
slide mechanics. Marine and Petroleum Geology, 22(1–2), 245–256. https://doi.org/10.1016/j.marpetgeo.2004.10.019
L’Heureux, J. S., Longva, O., Steiner, A., Hansen, L., Vardy, M. E., Vanneste, M., et al. (2012). Identification of weak layers and their role for
stability of slopes at Finneidfjord, Northern Norway. In Y. Yamada (Eds.) Submarine mass movements and their consequences. In Advances in
natural and technological hazards research (Vol. 31). Springer.
Lin, C. H., Kumagai, H., Ando, M., & Shin, T. C. (2010). Detection of landslides and submarine slumps using broadband seismic networks.
Geophysical Research Letters, 37(22), L22309. https://doi.org/10.1029/2010gl044685
Locat, J., Leroueil, S., Locat, A., & Lee, H. (2014). Weak layers: Their definition and classification from a geotechnical perspective. In Submarine
mass movements and their consequences (pp. 3–12). Springer.
Masson, D. G., Harbitz, C. B., Wynn, R. B., Pedersen, G., & Løvholt, F. (2006). Submarine landslides: Processes, triggers and hazard prediction.
Philosophical Transactions of the Royal Society A: Mathematical, Physical & Engineering Sciences, 364(1845), 2009–2039. https://doi.
org/10.1098/rsta.2006.1810
Micallef, A., Masson, D. G., Berndt, C., & Stow, D. A. V. (2007). Morphology and mechanics of submarine spreading: A case study from the
Storegga slide. Journal of Geophysical Research, 112(F3), F03023. https://doi.org/10.1029/2006jf000739
Morgenstern, N., & Price, V. (1965). The analysis of the stability of general slip surfaces. Géotechnique, 15(1), 79–93. https://doi.org/10.1680/
geot.1965.15.1.79
Norem, H., Locat, J., & Schieldrop, B. (1990). An approach to the physics and the modelling of submarine flowslides. Marine Georesources &
Geotechnology, 9(2), 93–111. https://doi.org/10.1080/10641199009388233
Puzrin, A. M., Germanovich, L. N., & Kim, S. (2004). Catastrophic failure of submerged slopes in normally consolidated sediments. Géotech-
nique, 54(10), 631–643. https://doi.org/10.1680/geot.54.10.631.56348
Puzrin, A. M., Gray, T. E., & Hill, A. J. (2015). Significance of the actual nonlinear slope geometry for catastrophic failure in submarine land-
slides. Proceedings of the Royal Society A: Mathematical, Physical & Engineering Sciences, 471(2175), 20140772. https://doi.org/10.1098/
rspa.2014.0772
Skempton, A. W. (1985). Residual strength of clays in landslides, folded strata and the laboratory. Géotechnique, 35(1), 3–18. https://doi.
org/10.1680/geot.1985.35.1.3
Somphong, C., Suppasri, A., Pakoksung, K., Nagasawa, T., Narita, Y., Tawatari, R., et al. (2022). Submarine landslide source modeling using the
3D slope stability analysis method for the 2018 Palu, Sulawesi, tsunami. Natural Hazards and Earth System Sciences, 22(3), 891–907. https://
doi.org/10.5194/nhess-22-891-2022
Spencer, E. (1967). A method of analysis of stability of embankments assuming parallel inter-slice forces. Géotechnique, 17(1), 11–26. https://
doi.org/10.1680/geot.1967.17.1.11
Sultan, N., Gaudin, M., Berne, S., Canals, M., Urgeles, R., & Lafuerza, S. (2007). Analysis of slope failures in submarine canyon heads: An
example from the gulf of lions. Journal of Geophysical Research, 112(F1), F01009. https://doi.org/10.1029/2005jf000408
Tappin, D. R., Matsumoto, T., Watts, P., Satake, K., McMurtry, G. M., Matsuyama, M., et al. (1999). Sediment slump likely caused 1998 Papua
New Guinea tsunami. Eos, Transactions American Geophysical Union, 80(30), 329–340. https://doi.org/10.1029/99eo00241
Watson, S. J., Mountjoy, J. J., & Crutchley, G. J. (2020). Tectonic and geomorphic controls on the distribution of submarine landslides across
active and passive margins, eastern New Zealand. Geological Society, 500(1), 477–494. https://doi.org/10.1144/sp500-2019-165
Zhang, W., Klein, B., Randolph, M. F., & Puzrin, A. M. (2021). Upslope failure mechanisms and criteria in submarine landslides: Shear
band propagation, slab failure and retrogression. Journal of Geophysical Research: Solid Earth, 126(9), e2021JB022041. https://doi.
org/10.1029/2021jb022041

Zhang, W., & Puzrin, A. M. (2021). Depth integrated modelling of submarine landslide evolution. Landslides, 18(9), 3063–3084. https://doi.
org/10.1007/s10346-021-01655-z
Zhang, W., & Randolph, M. F. (2020). A smoothed particle hydrodynamics modelling of soil–water mixing and resulting changes in average
strength. International Journal for Numerical and Analytical Methods in Geomechanics, 44(11), 1548–1569. https://doi.org/10.1002/nag.3077
Zhang, W., Randolph, M. F., Puzrin, A. M., & Wang, D. (2020). Criteria for planar shear band propagation in submarine landslide along weak
layers. Landslides, 17(4), 855–876. https://doi.org/10.1007/s10346-019-01310-8
Zhang, W., Wang, D., Randolph, M. F., & Puzrin, A. M. (2015). Catastrophic failure in planar landslides with a fully softened weak zone.
Géotechnique, 65(9), 755–769. https://doi.org/10.1680/geot14.p.218
Zhang, W., Wang, D., Randolph, M. F., & Puzrin, A. M. (2016). Dynamic propagation criteria for catastrophic failure in planar landslides. Inter-
national Journal for Numerical and Analytical Methods in Geomechanics, 40(17), 2312–2338. https://doi.org/10.1002/nag.2531

Zhang - How Small Slip Surfaces Evolve Into Large Submarine Landslides Insight From 3D Numerical

Uploaded by

Document Information

Copyright

Available Formats

Share this document

Share or Embed Document

Sharing Options

Did you find this document useful?

Is this content inappropriate?

Copyright:

Available Formats

Zhang - How Small Slip Surfaces Evolve Into Large Submarine Landslides Insight From 3D Numerical

Uploaded by

Copyright:

Available Formats

RESEARCH ARTICLE How Small Slip Surfaces Evolve Into Large Submarine

Received 9 FEB 2022

ZHANG AND PUZRIN 1 of 24

ZHANG AND PUZRIN 2 of 24

ZHANG AND PUZRIN 3 of 24

2.2. Governing Equations

Key assumptions for establishing governing equations are as follows.

ZHANG AND PUZRIN 4 of 24

Based on these assumptions, conservation of mass in each cell can be expressed by

𝜕𝜕𝜕 𝜕𝜕𝜕𝜕𝜕 𝜕𝜕𝜕𝜕𝜕

𝜕𝜕𝜕𝜕𝜕 𝜕𝜕𝜕𝜕𝜕2 𝜕𝜕𝜕𝜕𝜕𝑥𝑥 𝜕𝜕𝜕𝜕𝜕𝜕𝜕 𝜕𝜕𝜕𝜕𝜕𝑥𝑥𝑥𝑥 𝜏𝜏𝑤𝑤𝑤𝑤𝑤 + 𝜏𝜏𝑔𝑔𝑔𝑔𝑔 + 𝜏𝜏𝑑𝑑𝑑𝑑𝑑𝑑𝑑𝑑𝑑𝑑𝑑

𝜕𝜕𝜕𝜕𝜕 𝜕𝜕𝜕𝜕𝜕2 𝜕𝜕𝜕𝜕𝜕𝑦𝑦 𝜕𝜕𝜕𝜕𝜕𝜕𝜕 𝜕𝜕𝜕𝜕𝜕𝑥𝑥𝑥𝑥 𝜏𝜏𝑤𝑤𝑤𝑤𝑤 + 𝜏𝜏𝑔𝑔𝑔𝑔𝑔 + 𝜏𝜏𝑑𝑑𝑑𝑑𝑑𝑑𝑑𝑑𝑑𝑑𝑑

2.2.1. Marine Sediment Properties and Drag Force

where 𝐴𝐴f is the frictional drag coefficient,

ZHANG AND PUZRIN 5 of 24

ZHANG AND PUZRIN 6 of 24

3.1. Initiation of Slip Surface

where 𝐴𝐴 is a shape parameter,

ZHANG AND PUZRIN 7 of 24

Stability can be eventually achieved for the cases

ZHANG AND PUZRIN 8 of 24

3.3. Unstable Growth of Slip Surface

ZHANG AND PUZRIN 9 of 24

3.3.1. Slab Failure

ZHANG AND PUZRIN 10 of 24

3.4. Post-Failure Evolution and Redeposition of Slide Mass

ZHANG AND PUZRIN 11 of 24

4.2. Features of Dynamic Slip Surface Evolution

ZHANG AND PUZRIN 12 of 24

where 𝐴𝐴 is the Young's modulus

ZHANG AND PUZRIN 13 of 24

For plane strain and undrained conditions, the compression modulus,

4.3. Criteria for Unstable Growth of Slip Surface

𝑙𝑙𝑥𝑥 𝑙𝑙𝑦𝑦 Γ(1 + 1∕𝑛𝑛)Γ(1∕𝑛𝑛)

ZHANG AND PUZRIN 14 of 24

where 𝐴𝐴 is the shear stress ratio

ZHANG AND PUZRIN 15 of 24

𝐴𝐴 𝒍𝒍𝒙𝒙 (m) 𝐴𝐴 𝒍𝒍𝒚𝒚 (m) 𝐴𝐴 𝒏𝒏 𝐴𝐴 𝑨𝑨 (m )2

ZHANG AND PUZRIN 16 of 24

𝐴𝐴 𝒍𝒍𝒙𝒙 (m) 𝐴𝐴 𝒍𝒍𝒚𝒚 (m) 𝐴𝐴 𝒏𝒏 𝐴𝐴 𝑨𝑨 (m )

4.4. Effects of a 3D Slope Geometry

ZHANG AND PUZRIN 17 of 24

ZHANG AND PUZRIN 18 of 24

Appendix A: Numerical Scheme for 3D Modeling of Submarine Landslide Evolution

Conservation of mass in each cell can be expressed by

𝜕𝜕𝜕 𝜕𝜕𝜕𝜕𝜕 𝜕𝜕𝜕𝜕𝜕

𝜕𝜕𝜕𝜕𝜕 𝜕𝜕𝜕𝜕𝜕2 𝜕𝜕𝜕𝜕𝜕𝑥𝑥 𝜕𝜕𝜕𝜕𝜕𝜕𝜕 𝜕𝜕𝜕𝜕𝜕𝑥𝑥𝑥𝑥 𝜏𝜏𝑤𝑤𝑤𝑤𝑤 + 𝜏𝜏𝑔𝑔𝑔𝑔𝑔 + 𝜏𝜏𝑑𝑑𝑑𝑑𝑑𝑑𝑑𝑑𝑑𝑑𝑑

𝜕𝜕𝜕𝜕𝜕 𝜕𝜕𝜕𝜕𝜕2 𝜕𝜕𝜕𝜕𝜕𝑦𝑦 𝜕𝜕𝜕𝜕𝜕𝜕𝜕 𝜕𝜕𝜕𝜕𝜕𝑥𝑥𝑥𝑥 𝜏𝜏𝑤𝑤𝑤𝑤𝑤 + 𝜏𝜏𝑔𝑔𝑔𝑔𝑔 + 𝜏𝜏𝑑𝑑𝑑𝑑𝑑𝑑𝑑𝑑𝑑𝑑𝑑

ZHANG AND PUZRIN 19 of 24

A1. Stress Components at Cell Face Center

Similarly, the strain tensor

where 𝒆𝒆 is the deviatoric strain tensor

where 𝐴𝐴 is the shear modulus.

to the residual shear strength;

undrained shear strength.

A2. Shear Stress at Weak Layer

ZHANG AND PUZRIN 20 of 24

where 𝐴𝐴 is the shear stiffness.

A3. Finite Volume Scheme

ZHANG AND PUZRIN 21 of 24

At𝐴𝐴time 𝐴𝐴𝑛𝑛+1, the mass conservation Equation A1 is discretized as

ZHANG AND PUZRIN 22 of 24

Data Availability Statement

ZHANG AND PUZRIN 23 of 24